In utero exposure to antidepressants was recently associated with increased risk for autism in an analysis of a large health care claims database. This result is consistent with twin studies indicating greater concordance for autism among dizygotic twins compared to non-twin siblings, likely pointing to shared in utero risk. The need to clarify the risk posed by antidepressants is acute, as it has profound public health implications. Confirmation of the finding would support a rare modifiable risk factor for autism, and help to elucidate the underlying pathophysiology of the disorder. Conversely, if the finding is a false positive, it might lead to undertreatment of major depression among pregnant women, with consequent risks to both mother and fetus. The proposed study will utilize electronic medical record (EMR) data from more than 4 million individuals in a large New England health system. The investigators and their colleagues have previously applied an EMR query toolkit, which they helped to develop, to characterize treatment outcomes and conduct pharmacovigilance and genetic studies in psychiatry and beyond. Notably, this health care system includes large obstetrics, psychiatry, and pediatric practices, and is a referral center for child neuropsychiatri disorders. Children ages 4-12 with a diagnosis of autism who were delivered in this system will be identified using validated natural language processing tools. A matched (2:1) cohort of children with a diagnosis of ADHD will be identified, along with a matched (5:1) control cohort of children receiving routine pediatric care. The EMR will then be used to match each child with their mother, allowing characterization of maternal socio-demographic status, medical and psychiatric illness, and medication treatment during pregnancy. To further improve precision, Massachusetts state birth certificate data will be queried. This data includes details of perinatal care, exposures, and complications, as well as paternal age and socio-demographic status. Regression models will be applied to the resulting data set, encompassing more than 700 children with autism, 1400 children with ADHD, and 3500 healthy control children. In addition to examining the association between antidepressant exposure and autism liability, with appropriate control for confounding, this data set will allow investigation of other putative in utero and perinatal risk factors, creating a key resource for future studies. This project will innovate in two key respects. First, it will address the problem of confounding by indication present in prior antidepressant pharmacovigilance studies, using tools developed by the investigators for characterizing depression severity/comorbidity and course to better match cases and controls. Second, it will establish the utility of an EMR-based approach for studying pregnancy exposures and outcomes.